Back

Land use and climate warming interact to reshape global avian biodiversity through ecological filtering and homogenisation

Ding, C.; Newbold, T.; Etard, A.; Cornford, R.; Jung, M.; Brumm, K.; Jiang, Z.; Visconti, P.; Wang, Z.

2026-10-09 ecology
10.64898/2026.10.08.757631 bioRxiv
Show abstract

Land-use and climate change are dominant drivers of terrestrial biodiversity change, yet are usually analysed separately although they co-occur and may interact. We integrated global bird assemblage records from PREDICTS with temperature anomalies and species' ecological characteristics to test how land use modifies biodiversity responses to climate warming across three dimensions: local abundance and richness, compositional dissimilarity among assemblages, and species occurrence across ecological groups. Land use modified warming associations for both abundance and richness. Relative to primary vegetation, warming associations were more positive in low-intensity agriculture for both responses, but more negative for richness in high-intensity agriculture and urban land. Predicted richness remained below the reference, defined as primary vegetation at temperature anomaly 0, across all modified land uses, with the largest declines in high-intensity agriculture (-17.9%) and urban land (-21.8%). Within land-use classes, greater warming exposure was generally associated with higher compositional similarity, consistent with warming associated biotic homogenisation. Occurrence models revealed ecological filtering: broad-niched generalists were favoured in warmer modified habitats, whereas range-restricted habitat specialists were filtered out under intensive land use and warming. Conservation assessments should therefore integrate land-use-warming interactions with changes in assemblage composition and ecological strategies, particularly in agricultural and urban landscapes where warming further disadvantaged range-restricted specialists.

Matching journals

The top 2 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.